ROBUST CLASSIFIERS BY MIXED ADAPTATION

We show how to produce robust classifiers by combining supervised with unsupervised training. A supervised training phase exploits statistically scene invariant labeled data to produce an initial classifier. This is followed by an unsupervised training phase that exploits clustering properties of unlabeled data. We refer to this two-phase process as mixed adaptation. We present a probabilistic model supporting this technique. We present synthetic and real examples illustrating mixed adaptation. These examples include the detection of unspecified dotted curves in dotted noise and the detection and classification of vehicles in cinematic sequences of infrared imagery.